Real-space post-processing correction of thermal drift and piezoelectric actuator nonlinearities in scanning tunneling microscope images

TL;DR

This paper introduces a real-space correction method called DHCT for thermal drift and piezoelectric nonlinearities in STM images, improving image accuracy by using internal structural standards and a distortion model.

Contribution

The paper presents a novel real-space correction technique for STM images that accounts for thermal drift and piezoelectric nonlinearities using internal standards and a comprehensive distortion model.

Findings

Effective correction of STM images demonstrated with graphite(0001)

Enhanced image analysis capabilities enabled by correction

Quantitative assessment of image noise and feature confidence

Abstract

We have developed a real-space method to correct distortion due to thermal drift and piezoelectric actuator nonlinearities on scanning tunneling microscope images using Matlab. The method uses the known structures typically present in high-resolution atomic and molecularly-resolved images as an internal standard. Each image feature (atom or molecule) is first identified in the image. The locations of each feature's nearest neighbors (NNs) are used to measure the local distortion at that location. The local distortion map across the image is simultaneously fit to our distortion model, which includes thermal drift in addition to piezoelectric actuator hysteresis and creep. The image coordinates of the features and image pixels are corrected using an inverse transform from the distortion model. We call this technique the thermal-drift, hysteresis, and creep transform (DHCT). Performing the correction in real space allows defects, domain boundaries, and step edges to be excluded with a spatial mask. Additional real-space image analyses are now possible with these corrected images. Using graphite(0001) as model system, we show lattice fitting to the corrected image, averaged unit cell images, and symmetry-averaged unit cell images. Statistical analysis of the distribution of the image features around their best-fit lattice sites measures the aggregate noise in the image, which can be expressed as feature confidence ellipsoids.